Repository navigation
Two lectures both read 工作搜寻 V — the five myst#760 renumberings were never delivered; two refactor backports also outstanding #266
Description
The 2026-08-20 full-corpus audit (project-translation#50) content-verified this edition's drift. With mccall_risk now seeded (#264), what remains is small and precise — but the seed has turned the missed renumberings into an active collision:
The Job Search numbering (reader-facing)
myst#760 inserted Job Search V and shifted five siblings. The shifts were in auto-closed #242 (the action-translation#280 race) and never arrived by any other route. Now that #264 has correctly delivered mccall_risk as 工作搜寻 V, the edition has two adjacent lectures both numbered V (mccall_risk and mccall_persist_trans, _toc.yml rows 81–82), and everything after is one behind:
| Lecture | H1 today | Should be |
|---|---|---|
mccall_persist_trans |
工作搜寻 V | VI |
career |
工作搜寻 VI | VII |
jv |
工作搜寻 VII | VIII |
odu |
工作搜寻 VIII | IX |
mccall_q |
工作搜寻 IX | X |
Five H1 lines (plus each file's translation.title and, for jv, the embedded {index} target). Note mccall_persist_trans's state file claims the change is synced — its 2026-08-18 UPDATE wrote a source-sha past the skipped hunk (filed as action-translation#287) — so this will not surface as OUTDATED in translate status.
Two small refactor backports, same lost window
ak_aiyagari— myst#706 (+9/−9):hh→householdinside the jitted helper bodies; target still has 23hh.j_grid-style lines. The audit's reviewer also noted the current form silently binds the module-globalhhinside functions that take ahouseholdparameter — inert today, wrong if reused.harrison_kreps— myst#669 (+7/−4):price_single_beliefsrefactor (imbq_inv→ solve-basedA), typo fixes, bold→italic;imbq_invstill present ×2.
And one state-file-only touch
likelihood_bayes is the mirror case: content already carries myst#977 (simulate(a, b, rng, …) present, prange gone) but .translate/state points at the pre-#977 SHA. Left alone, a future forward run may needlessly rewrite a current translation — refresh the state file only.
All verified against source 285a1c6 / target b136ab3; per-lecture evidence in the audit's drift table.
Activity
mmcky commentedon Aug 24, 2026
Now live and reader-visible — confirmed on the published site, 2026-08-24.
Publishing
mccall_riskin publish-2026aug24 took this collision from latent to reader-facing. Both lectures now serve the same numeral in their H1 and their browser-tab title:mccall_risk.htmlmccall_persist_trans.htmlBecause the numerals are in the source H1s, the duplication also appears in the sidebar of every page that renders the Job Search section, not just on the two lectures themselves.
The full picture against the English source confirms this issue's premise exactly — the zh-cn numerals run one behind from
mccall_persist_transonward:mccall_riskmccall_persist_transcareerjvodumccall_qOne rider for whoever does the renumbering:
jv's numeral also appears inside its{index}role, where the English half of the entry already reads "Job Search VII" and needs to move to VIII in the same pass. Worth grepping for in-prose cross-references that name these lectures by numeral too — there are link texts reading 工作搜寻 IV onmccall_persist_transandmccall_risk.Note this is one of two numbering defects on the same two pages, and they have different causes. This issue is the Roman numeral (undelivered myst#760 renumberings). Separately, both pages also serve the same chapter number
57., because the publish build deployed stale HTML for every page it did not rebuild — filed as #270. Fixing this issue will not fix the57.duplication, and vice versa.mmcky commentedon Aug 24, 2026
The five renumberings are done — #273. This issue stays open for the rest.
#273 shifts all five, matching the English source verbatim (each numeral checked file by file against
QuantEcon/lecture-python.myst):mccall_persist_transV→VI,careerVI→VII,jvVII→VIII,oduVIII→IX,mccall_qIX→X. Eleven lines in five files — each#heading and itstranslation.title, plus the English half ofjv's{index}target, which also read VII and would have left the index entry on the old numeral.Job Search I–IV were deliberately left alone, since they sit before the insertion point. That keeps the two in-prose cross-references in
mccall_persist_trans(工作搜寻 III and 工作搜寻 IV) correct. A full-corpus sweep for工作搜寻 <numeral>andJob Search <numeral>confirms the series now reads I through X with no duplicate and no gap.Still open here, deliberately not bundled, because both touch code cells and so need their own execution run rather than riding tonight's prose-only publish:
ak_aiyagari— the myst#706 backport (hh→householdin the jitted helper bodies; 23hh.j_grid-style lines remain), including the module-global binding the audit's reviewer flagged.harrison_kreps— the myst#669 backport (price_single_beliefsrefactor,imbq_inv→ solve-basedA, still present ×2, plus the typo and emphasis fixes).likelihood_bayes— the state-file-only refresh; content already carries myst#977, only.translate/statepoints at the pre-#977 SHA.Note the state-file trap recorded in the body still applies to item 3 and to anything touching
.translate/state: the sync path rebuilds those files from its known fields, so an unknown key does not survive — that is action-translation#290.mmcky commentedon Aug 24, 2026
Correction to my earlier comment on this issue. Where it said the publish build "deployed stale HTML for every page it did not rebuild", that theory was investigated and found wrong. All 145 pages were rebuilt from the tagged sources; what was stale is the Sphinx environment restored from the cache artifact, so fresh pages were rendered against a toctree predating
mccall_risk. The reader-visible symptoms are unchanged and #270 carries the corrected account. Nothing about this issue's own scope changes.Renumber Job Search VI-X, and point baseurl at the live site (#273)
mmcky commentedon Aug 24, 2026
Reopened — this issue was closed by accident when #273 merged, and my PR body caused it.
#273 delivers only the five renumberings. The rest of this issue is untouched and still needs doing: the
ak_aiyagaribackport (myst#706), theharrison_krepsbackport (myst#669), and thelikelihood_bayesstate-file refresh.The cause is worth recording because it is a trap with a sharp edge. My PR body contained the sentence "This does not close #266, which also carries the two refactor backports" — written specifically to prevent an auto-close. GitHub's auto-linker does not parse grammar: it saw
close #266and closed it. The negation made no difference.The merge commit on
mainis safe — it reads "Refs #266 …remain open there" — so this will not recur on that commit. I have edited the PR body to remove the phrase.Nothing about the work here changed; only the issue state, which is now corrected.